Gallery of master's theses

2020

  • Mgr. Laura Pauline Gschwandtner - Deep learning for detecting interictal EEG biomarkers to assist differential epilepsy diagnosis
    ... more
    This project explores novel approaches to assessing electrical brain activity recorded through electroencephalography (EEG) for epilepsy diagnosis. The thesis at hand presents a machine learning algorithm for automatic classification of EEG recordings into epileptic and non-epileptic. In epilepsy patients, relatively short EEG recording periods without apparent epileptic events like seizures usually still show patterns different from non-epileptic EEG activity. Such interictal epileptiform discharges (IEDs) can serve as EEG biomarkers. However, to date no reliable and unambiguous diagnostic analysis solely based on the known IEDs has been established. Identifying additional interictal biomarkers for diagnosis could contribute to faster procedures and help to prevent misdiagnoses and ineffective or harmful treatment. Faster diagnostic procedures could furthermore save essential resources and make hospital beds available in urgent situations. The presented thesis tackles this issue by making an end-to-end machine learning algorithm search for distinct patterns in EEG data and learn to use them to differentiate epileptic from non-epileptic recordings. By exploring means of explainable AI (XAI), the goal was to make the detected patterns explicit. Creating such explanations for algorithmic decisions further carries high ethical relevance in the health care field. Practitioners will only gain trust in a system and agree to use it, if the automated decisions are logical and the reasoning process can be explained clearly. This level of interpretability is not inherent in the deep neural network algorithms applied. Thus, within the scope of this work the concept of explainable AI was explored and implementation was attempted. The presented study shows that automatic EEG analysis using deep learning is feasible. At the same time, extensive further research will be necessary to create clinically applicable, highly accurate and transparent algorithms for epilepsy diagnosis from routine EEG.

  • Mgr. Johanna Köllner - The City as a Playground: Influence of Practicing Parkour on Divergent Thinking
    ... more
    According to the 4E approach to cognition, cognition does not only emerge from the brain, but rather arises from a cooperation between brain, body, and the environment. Hence, possible ways of interaction between humans and their environment, so-called affordances, play a crucial role in a number of cognitive processes. Most spaces designed for a single purpose offer only a few affordances. Such tight spaces are perceived as less vivid in comparison to so-called loose spaces, offering their users a large variety of affordances. The attitude towards the environment is often challenged in a sporting activity, which is called parkour: its practitioners, so-called traceurs, exercise the ability to loosen up spaces by finding additional affordances in initially mono-functional surroundings, as they use urban structures creatively. The aim of my thesis is to approach an interdisciplinary understanding of the importance of human-environment interactions. Therefore, I hypothesized that parkour training had a beneficial effect on divergent thinking (DT), as an aspect of cognition. DT, being a type of creative thinking, describes the ability to generate a range of diverse ideas as possible solutions to a single problem. Traceurs might apply a similar type of thinking when loosening up urban environments. In my experiment, a group of traceurs solved two DT tasks before and after a parkour training session in an unfamiliar environment. As a control group, I chose gymnasts who trained on gymnastics equipment instead. The DT tasks of both groups were evaluated using a new computational scoring method based on semantic distance. The participants’ DT performance differed depending on the type of task: whereas one task indicated an increased number of ideas after training, the other showed decreased originality across all participants. Furthermore, traceurs had a higher number of ideas that were more original than the gymnasts’. Over all participants, DT performance before and after the training session correlated positively, indicating that my scoring method showed acceptable feasibility. Although the results are inconsistent, I succeeded in providing a preliminary explorative approach towards the connection between 4E cognition, creativity, the concept of affordances, and urban design. This thesis introduces a new paradigm to investigate how physical activity and interaction with the environment might affect cognitive abilities.

  • Mgr. Rebecca Rose Lebens - Regular and irregular morphological inflection in bilinguals: evidence from Huntington’s disease studies
    ... more
    In the field of neurolinguistics, one area of interest is the basal ganglia. Research into the linguistic impairments caused by Huntington’s disease has produced several theories regarding the neurological underpinnings of morphological inflection and the role of the basal ganglia in such processes [Ullman et al 1997, Kargieman et al 2014]. From the existing studies, there are contradictory results leading to differing positions on the necessity of the basal ganglia in lexical retrieval and grammatical rule processing [Teichmann et al 2005, Longworth et al 2005]. The main two arguments debate whether the basal ganglia are necessary for grammatical rule-based tasks or if they are involved in the general inhibition of competing lexical forms. In order to test these theories, tasks of morphological inflection in which there is a dissociation between grammatical processing and lexical retrieval, such as the comparison of regular and irregular English past tense conjugation, are used [Ullman et al 1997, Tyler 2002, Longworth et al 2005]. As performance in inhibitory tasks is stronger in bilinguals, the same research in bilinguals could provide a context for comparing these two hypotheses.

  • Mgr. Oswaldo Macedo - Recognizing and Imitating Emotions in a Robotic System
    ... more
    For decades, interest in emotion recognition by artificial intelligence has been growing steadily. Lately, especially because of improvements in the processing capabilities of computers and the creation of new machine learning algorithms, this path of research grew even faster, substantially improving the accuracy of the task. Currently, emotion recognition through facial expressions can achieve close to 73% accuracy with state-of-the-art implementations tested with images in natural settings (Pramerdorfer & Kampel, 2016). Comparably, human accuracy tested on the same images reaches 65±5% (Goodfellow et al., 2013). This thesis consists of two main parts: first, a neural network based emotion recognition system trained with static images of faces with a neutral expression and six basic emotions (sadness, happiness, anger, fear, disgust, surprise); and second, an emotion imitation system trained on outputs of the emotion recognition system, learning to imitate the facial expressions of new images. The thesis combines state-of-the-art level facial expression recognition implementations with an emotion imitation system to produce a human–robot interactive system capable of perceiving and imitating human emotions through facial expressions. For this purpose, state-of-the-art neural network models are modified and trained with static images of faces from three commonly used datasets of facial expressions. Also, the imitation system is trained with associative learning, using the resultant information from the previously trained neural network. The emotion recognition system achieves an average accuracy of 62% on the most challenging, natural setting dataset used in this research; while achieving an average accuracy of 83% on laboratory setting datasets. The emotion imitation system correctly associates all images with an average accuracy of 96% on the laboratory setting dataset, which results in correctly imitating emotion with a 78% accuracy in combination with the emotion recognition system. Regarding the natural setting dataset, association accuracy is 91% on average, which combined with the recognition system results in 56% average imitation accuracy. The results show that the robotic system would not be very accurate at imitating emotions in a natural setting with current emotion recognition capabilities, nevertheless, on a laboratory setting, it could be viable.

    Keywords: associative learning, computer vision, convolutional neural networks, emotion, facial expression recognition
  • Mgr. Kristína Miklošová - Saliency models analysis for paintings
    ... more
    The topic of visual saliency spreads across numerous disciplines. In this work we focus on computational saliency models and their performance in the context of art. We examine how saliency models perform on images with specific characteristics. We explore a saliency of 14 different digitized paintings, all representing a biblical scene of The Last Supper. We process the data and evaluate the performance of four saliency models against them. Two of the models are using traditional approach and two are based on deep neural networks. For the evaluation we use AUC, NSS, and CC metric. As a ground truth we use eye-tracking data from 35 participants. Moreover, we incorporate face detection algorithm to one of the models and slightly improve its performance. Our analysis shows that deep-learning models predict the most salient parts of the paintings closest to real eye fixations.

  • Mgr. Peter Šebáň - Kea and the ephemeral reward task: success and hurdles explored
    ... more
    New Zealand parrot species kea (Nestor notabilis) and their cognitive abilities are at the center of our research. Kea are well-known for their curiosity, complex social interactions and problem-solving. Their cognitive abilities were investigated in two different experimental tasks, both of them in two different settings. In the ‘Two versus one quantity discrimination task’, it was revealed that kea are able to discriminate between one piece of reward and two pieces of reward, however their performance was not as straightforward as expected. The main objective of our research was to explore the performance of kea on the Ephemeral reward task. It was revealed that despite difficulties, kea are able to solve this task. This intriguing task was solved only by a small number of other species. The general principle of the Ephemeral reward task is as follows: There is a permanent reward that can be chosen at any time, and an ephemeral reward which is removed, if not chosen first. Choosing the ephemeral reward first allows the choice of the permanent reward after, resulting in two rewards per trial. Choosing the permanent reward first, leads to the end of a trial and the gain of only one reward. Previous studies of the Ephemeral reward task on other species directed their focus on generating relevant ecological conditions of given species and on their impulsivity control. Our research also followed this direction and for that reason, two different experimental settings were employed.

    Keywords: animal cognition, kea (Nestor notabilis), Ephemeral reward task, ecological relevance, decision making
  • Mgr. Igor Slovák - Few-shot face recognition using artificial neural networks
    ... more
    Face recognition is an essential part of visual perception. Few-shot learning is inspired by the unique ability of humans to recognize objects only after one or a few presentations. We adopt an interdisciplinary approach by combining the research from neuroscience for face perception as well as computer science research for using neural networks in the face perception domain. We focused on the implementation of few-shot learning models for face recognition, namely the Prototypical networks and the Siamese networks. We have used Labeled Faces in the Wild dataset (Huang et al., 2007). We provide an analysis showing that few-shot learning models do not provide sufficient classification accuracy compared to humans. However, with more computational power allowing a more thorough exploration of the deep learning models for few-shot face recognition is a very promising area.

  • Mgr. Adam Štefunko - Testing Independence of Compte's Spatial Working Memory Model from the Neuron Model
    ... more
    Spatial working memory, the ability to memorize locations for a short period of time, has been studied for decades. Compte and his colleagues proposed a model of the spatial working memory used to simulate an oculomotor delayed-response task. The model consisted of excitatory pyramidal cells and inhibitory interneurons. Pyramidal cells were spatially distributed according to their sensitivity to different peripheral cue angles. Pyramidal cells closer to each other had stronger synaptic connections than those farther apart. The oculomotor delayed-response task was simulated as stimulation of pyramidal cells with a selective transient current. Pyramidal cells whose preferred angle was close to the presented cue angle exhibited elevated spiking activity which persisted after the disappearance of the stimulus. Leaky integrate-and-fire neurons were employed in the model. We implemented the model using the Simple model of spiking neurons described by Izhikevich and tested how this model behaves when different model parameters are manipulated. We showed patterns appearing in the region where neurons with preferred cue angle close to the presented cue angle resided. Although these patterns were not completely identical with findings of Compte and his colleagues, they represented a prominent and distinct manifestation of the spatially tuned stimulation of the neurons during the cue presentation. Therefore, we managed to simulate the oculomotor delayed-response task using the Simple model of spiking neurons. Results of our thesis may be used in replicating existing computational studies related to this task using the simple spiking neurons and integrating the task and the spatial working memory model to the simple spiking neurons related research.

    Keywords: simple model of spiking neurons, spatial working memory, synaptic connection
  • Mgr. Nicole Vella - A quantitative analysis and characterization of mouse CA1 neuron classes: A knowledge base for computational models
    ... more
    In an age where computational power and technological advances have allowed us the possibility to simulate whole brain regions, the data necessary for such models is in high demand. Currently available experimental data, however, is both sparse and non-uniform, due to the various sources it is extracted from and the different methods and technologies used for its procurement. The mouse hippocampus CA1 region is one of the most anatomically studied brain structures due to its relative simplicity and availability of transgenic mice. The structure itself is a window into the neural bases of cognition; representing the processing, storage and emotional flavouring of our episodic memories. A complete anatomical model, down to the cellular level of the mouse hippocampus, however, has not yet been simulated, with data scarcity being one of the main culprits for the delay. In this study, therefore, through an extensive literature review and data mining, the constituent glutamatergic and GABAergic neurons of the CA1 region of the mouse hippocampus were classified and quantified. By means of extensive calculations, the data has been streamlined and presented with laminar dorso-ventral accuracy. Such a quantitative knowledge-base is not only a requirement for data-driven, large-scale computational simulations but also acts to inspire and focus future experimental projects onto current gaps in the data thus aiming towards a more holistic histological profile of the mouse CA1 hippocampal region.

2019

  • Mgr. Ondrej Hadidom - The role of the left inferior frontal cortex in semantic retrieval: a tDCS study
    ... more
    Our general knowledge about the world is stored in semantic memory in form of conceptual representations. In order to efficiently use these concepts, employment of various semantic retrieval mechanisms is required. Context can provide cues automatically and effortlessly eliciting semantic memories, however in case of their absence or insufficiency we need to devote our cognitive resources to controlled semantic retrieval. Cognitive control encompasses aspect as inhibition of prepotent habitual responses and switching providing us with cognitive flexibility in task-dependent context. Left lateral inferior frontal cortex (LIFC) has been proposed as vital neural substrate responsible for language production processing and controlled semantic retrieval. Transcranially delivered neurostimulation of left LIFC via low-intensity direct current (tDCS) has been also suggested to enhance performance in lexical-semantic task. However, effect of tDCS over LIFC on specific pre-retrieval and post-retrieval semantic processing remains to be undescribed. In this project we applied anodal tDCS over LIFC and administered tasks allowing us to distinguish between specific aspects of semantic retrieval (categoric verbal fluency, associative words production, dissociative words production, production of words without prime). We tested participant (N=27) in within subject design consisting of three consecutive sessions devided by least 5 days. Analysed data suggestsed that tDCS applied over left LIFC may improve controlled semantic processing, as assessed by categoric verbal fluency and dissociative word production measures. According to our results we conclude that used neurostimulation resulted in boosted controlled inhibition of prepotent automatic responses.

    Keywords: controlled semantic retrieval, transcranial direct current stimulation, left lateral inferior frontal cortex
  • Mgr. Endre Hamerlik - Recurrent neural network model of phonological development using distributed representations
    ... more
    The goal of the presented thesis is to build a psycholinguistically plausible computational model of phonological development, inspired by the existing models of Takac, Knott, and Stokes, 2017 and Dell, Juliano, and Govindjee (1993). Takac’s current model of phonological development is considering localistic representations of phonemes and word-meanings (one-hot vectors, moretechnically). Moreover, according to their study, even the localistic neural ac-tivations provide an explanation for several patterns shown by infants duringphonological development. Merely all the neurobiological or even neuro-imaging studies claim, thatneural activations are distributed within the brain; Moreover, neither an area, where an activation pattern is exceeding during a given treatment, can be considered as the brain area responsible for phenomenon invoking the activation. Suggesting, that every brain area respond to different treatments to differentextent. (Tremblay and Dick, 2016). Our thesis aims to reproduce this phe-nomenon within the hidden layer of our Artificial Neural Network. Thus, in the experimental part of our work, we developed a Simple Recur-rent Neural Network (SRN) model, trained under circumstances of the sourcestudy by Takac, Knott, and Stokes (2017). However, in our study, phoneme and even the meaning representations are modified in a neurobiologically more plausible way. The model is tested for performance during the training, analogically to children learning their first 300 words, approximately in their first two years. Nevertheless, the methodology of analyzing infant’s datawill be maintained, accounting for phonological Neighborhood Density and related phenomena. Our results suggest, that the phonological Neighborhood Density effect is not necessarily related to localness of representations. In line with our novel findings, we provide an alternative explanation of the present phenomenon. Furthermore, speech error analysis (in line with Dell, Juliano, and Govindjee,1993) is presented on the speech errors produced by our model, in order to support our hypotheses based on the Parallel Distributed Processing Paradigm.

  • Mgr. Richard Leckéši - Neurophysiology of emotion transfer between professional dancers and viewers of dance using near-infrared spectroscopy
    ... more
    Recent studies have shown that an individual has an ability to routinely pick up and even feel emotions transmitted via several artistic media like music, film or opera. However, there is little empirical data on the medium of dance. Which is quite surprising due to dance being considered one of the most direct forms of, nonverbally communicating an emotion, although it is not known how well people can use movement to transmit an emotion and if the viewers respond similarly or correctly. In the theoretical part we focus on defining emotions, emotion transfer, cognitive an affective empathy and dance as a form of communication and as an aesthetic movement from philosophical, psychological and neurophysiological point of view. In the experimental part we try to determine if there are any patterns or correlations between the medial prefrontal cortex, inferior frontal gyrus and bilateral temporal parietal junction areas of the brain of a professional dancer while in attempt to communicate an emotion via dance and the evaluation of the performance by an audience. Our thesis consists of two experiments. In the first experiment 10 professional dancers were asked to communicate a specific emotion via movement while their brain activity was being measured by near infra-red spectroscopy and after each performance they were asked several questions concerning their dance. In the second experiment recordings of the dancer´s performance was shown to 15 non-dancer observers. Their task was to correctly guess the emotion being performed and answer a few questions concerning the performance. After data analysis no statistically, significant correlation was found between the brain activation of the dancer and the evaluation of the observers. However, correlations were found between the specific brain areas of our focus suggesting an interconnected system. Furthermore, correlations were also found between the evaluated expressiveness of the dance by the audience and their general liking of the performance and confidence of correct guessing of an emotion. Also, significant percentage of correct guesses of emotions by observers have been found suggesting an existence of information transfer between the observer and the dancer.

    Keywords: emotion, emotion transfer, emotion contagion, dance, aesthetics, communication, near infra-red spectroscopy, medial prefrontal cortex, inferior frontal gyrus, temporal parietal junction
  • Mgr. Xenia-Daniela Poslon - Vicarious Sensorimotor Activation and Dispositional Empathy
    ... more
    Embodied approach to social cognition suggests that the central aspect to the understanding of others relies on the activation of neural structures involved in our own personally experienced actions, bodily states, and emotions. Vicarious sensorimotor resonance, i.e. brain activation resulting from observing other people’s actions or sensations, is believed to contribute to the neural underpinnings of empathy. Two electroencephalography (EEG) studies measuring resonant responses of sensorimotor cortex to the pain and touch of others were conducted to further explore the associations between the vicarious oscillatory activity and individual differences in empathy. The aim was to assess the relationship between event-related desynchronization (ERD) and synchronization (ERS) of sensorimotor mu and beta rhythms and behavioral measures of dispositional empathy. In study 1, participants (N = 37) observed videos showing hands being either penetrated with needle or touched with a cotton swab. The stimuli were presented on a monitor placed in front of the participants. In study 2 (N = 30), identical stimuli were used, but were presented horizontally on a monitor placed over the participant's hand, in order to increase bodily self-attribution of the target hand. In both studies, participants filled out the Interpersonal Reactivity Index (IRI) questionnaire, measuring four dimensions of empathy: Perspective Taking, Fantasy scale, Empathic Concern, and Personal Distress. Event-related spectral power modulation (ERD/ERS) was assessed for each subject within the frequency bands of 7-12 Hz (mu) and 13-30 Hz (beta) over sensorimotor cortex. The results showed significant mu and beta ERD following the perception of stimuli. Even though studies involved different setup, we didn’t find any statistically significant differences in the perception of stimuli between these two conditions. Finally, we observed weak associations between certain IRI subscales and mu and beta ERD evoked with the perception of painful hand treatment. Mu ERD was predicted by the participants’ self-reported tendency to experience distress when witnessing the suffering of others, while individuals with higher tendency to feel empathic concern for others also showed stronger beta ERD. Possible interpretation of these results is discussed, as well as several directions for future research.

  • Mgr. Márius Rak - Application of cognitive science on improving the process of user onboarding into a new software interface
    ... more
    User’s beginnings with software tools can be difficult. A user might need help during a transition phase from novice to advanced user. This process and accompanying help are known as onboarding. Ease of the transition phase and the effectivity of the onboarding can influence the acceptance of the software. Field of human-computer interaction (HCI) currently uses techniques that could be applied for evaluation of onboardings. However, these techniques have drawbacks, such as high subjectivity, lower ecological validity and higher time demands. Two new methods employing eye-tracking, electroencephalography (EEG) and their combination are proposed. The methods are based on areas of interest (AOI), event-related potentials (ERP) and their fusion eye-fixation related potential (EFRP).

  • Mgr. Marek Sokol - Influence of working memory load and task importance on prospective memory performance
    ... more
    Prospective memory is special type of memory, oriented into future intentions. This type of memory enable humans to function effectively within the environment. However, prospective memory is prone to errors and can fail under various circumstances. Theoretical part unfolded the nature and of prospective memory using knowledge from psychological experiments, neuroimaging studies, psychophysiological experiments and also computational modelling approach. Experimental part tried to explore the unknown influence of interplay between factors of working memory load and task importance on prospective memory performance. The performance measure was meant to reflect the naturalistic way of assessing prospective memory. Our sample (N = 55) was separated by working memory load and task importance manipulation with 2x2 design, resulting in total of 4 groups. At the beginning of experimental procedure, participants were told to remind experimenter of completing form of participation after the experiment, which was our real performance measure. Also task importance manipulation happened there, when experimenter put different emphasis on reminding of the form. Participants then performed working memory digit span task as a secondary task where working memory load manipulation was employed. After evaluation of our data, the results showed no significant difference in interplay of working memory load and task importance factors. However, we observed significance of importance manipulation factor alone, this was not a concern of our thesis. Nevertheless, we provided possible interpretation of our results and discussed possible gains from our thesis to prospective memory research.

  • Mgr. Romana Umrianová - Role of theta oscillations in prefrontal cortex in semantic retrieval: tACS study
    ... more
    Neural oscillations are crucially involved in neuronal communication and computations that are necessary for complex cognitive functions and adaptive behaviour. Semantic cognition is a fundamental system that enables semantic memory retrieval. Recent studies have demonstrated that neuronal theta oscillations in prefrontal perisylvian brain regions may play a pivotal role in binding of semantic representations. However, theta oscillations have been also associated with the involvement of cognitive control and working memory functioning which putatively support controlled semantic processing. Thus, the functional role of theta oscillations in semantic retrieval remains poorly understood. For this purpose, transcranial alternating current stimulation (tACS), an external frequency-specific modulator of endogenous cortical fluctuations, was used to modulate endogenous theta band oscillations. In the present study, we applied active and sham tACS over the left inferior frontal cortex and the contralateral supraorbitary region to entrain prefrontal theta oscillations at 6Hz (θ-tACS) over the inferior frontal cortex in 27 healthy participants. The tACS conditions were delivered in separate sessions using a pseudo-randomized and properly balanced cross-over experimental design. Participants completed tasks assessing automatic and controlled retrieval performance in three blocks within each session before, during and immediately after the stimulation. Our findings indicate that θ-tACS significantly facilitated retrieval tasks involving automatic processes (i.e., delivering free and unconstrained associations) but impaired controlled retrieval that required cognitive control (i.e., the inhibition of habitual responses and switching between semantic sets). Our study provides an important experimental evidence indicating that neuronal theta oscillation may constitute a neurocognitive mechanism for semantic binding, rather than cognitive control. We conclude that theta oscillations over left prefrontal cortex may support well-established semantic connections or strengthen their metastable activation, which enhances fluent retrieval. These conclusions should be supported by further empirical research.

top

2018

  • Mgr. Andrej Brinkač - Application of Physiological Measures for User Experience Testing: A Skin Conductance Experimental Study
    ... more
    With user-oriented web and application development becoming an ever-greater source of competitive advantage, the concept of user experience has gained wide attention over the past decade not only among practitioners, but also among academic researchers. The need to have measured data that is as accurate as possible arises, in order to aid in the process of creation of more user-friendly interfaces. While most studies resorts to ‘typical’ methods of evaluating an interface’s usability, via measures such as completion time, success rate, or questionnaires, the subjective nature of the methods generates concerns as to the reliability of the data. Consequently, this thesis aims to contribute to the limited literature concerning new, more objective and efficient metrics for usability testing, by proposing that physiological response measures can be applied to the field of user experience. We combine typical task-objective measures (task success rate, task completion time) and subjective participant evaluation (task difficulty evaluation, user experience questionnaire) with the physiological approach of skin conductivity. Our main objective within this study is hence to analyze the relationship between physiological measures and traditional usability measures. In order to evaluate the relationships between these measures, we design an experiment with a subsequent correlation analysis. We investigate whether skin conductance response is effective in identifying specific emotions in users’ interactions with a web application, and examine if skin conductivity measures can enrich or support traditional usability metrics.

    Keywords: UX, User Experience, Usability testing, Skin conductance, Human computer interaction, physiology
  • Mgr. Miroslava Galasová - Who Will Save Innocent Kitties with Counterfactual Priming?
    ... more
    How can you save an innocent kitty? Easy – do not make dumb decisions. However, how can we avoid dumb decisions when cognitive biases lurk to prevail our rationality? In this case, counterfactual priming can be the right choice. Cognitive biases create barriers to optimal choices and can lead to serious consequences. Therefore, we should try to reduce them. There are several debiasing strategies available. We decided for counterfactual priming that seems to be the promising one. In previous studies, counterfactual priming was created by scenarios in which the main character was an unfamiliar person. However, neuroscientific evidence suggests that when we imagine “if only” situations in which we are the main actors (e.g. “If only I was five minutes earlier.”), different brain regions are activated than when we imagine “if only” scenarios with someone else. Thus, we assumed that counterfactual priming with self-based scenario will have different effect on reducing cognitive biases than other-based scenario. We conducted a between subject experiment (N = 266) in which participants solved a task either on attribution error, confirmation bias, or sunk cost fallacy. Then, they were primed by self-based or other-based counterfactual scenarios, and after the priming, they solved one of the two remaining tasks. Moreover, we were interested whether actively open-minded thinking (AOT) and age category (under and above 25) affect susceptibility to the three cognitive biases. We found out that counterfactual priming reduced attribution error; however, only priming by “self” led to a significant difference compared to the control group. Priming by “others”, on the other hand, slightly reduced confirmation bias. Sunk cost fallacy was not reduced by counterfactual priming at all. Nevertheless, we observed a moderation effect of AOT and age category on the intervention. Counterfactual priming in general and counterfactual priming by “self” reduced attribution error among people who reached medium and high scores in dogmatism. In contrast, priming by “others” affected only younger participants. Among older participants this type of counterfactual priming led to higher susceptibility to attribution error and sunk cost fallacy than priming by “self”. We conclude that the effect of counterfactual priming on certain cognitive biases depends on the character in priming scenarios, cognitive capabilities (such as AOT), and age of participants. These findings can be used in proposing further practical interventions to optimize individual and group decisions … and for saving innocent kitties.

  • Mgr. Lucia Hrašková - Activation of the Mirror Neuron System by Emotional Facial Expressions: an fMRI Study
    ... more
    The ability to make inferences about the mental and emotional state of others, partly enabled by reading emotional facial expressions, is crucial to social interaction. Mirror neurons, a class of neurons active both when observing and experiencing a motor act, have been argued to be its underlying mechanism. The goal of the present work is to deepen the understanding of this mechanism by identifying brain regions that are activated by observing emotional facial expressions, and potentially part of the mirror neuron system, using functional magnetic resonance imaging (fMRI). Following a thorough review of existing literature on emotions, empathy, and the mirror neuron system, as well as similar studies on which this work elaborates, we describe the technique of fMRI, before proceeding to the analysis itself.The results suggest that the brain regions activated by our experiment include some of the key areas commonly considered to be part of the mirror neuron system - BA44, BA45, the inferior parietal lobule, the insula, the amygdala, and the inferior frontal gyrus - and confirm that emotions with higher subjective arousal are more easily induced by films designed to elicit them. The work aims to be transparent about its underlying assumptions, limitations, and the decisions made over the course of it. The latter are described in considerable detail, with a reflection on the process itself constituting an integral part of it. The detailed description of fMRI analysis using FSL also makes it potentially useful as a tutorial for beginners.

    Keywords: emotions, facial expressions, fMRI, mirror neurons
  • Mgr. Anton Kováč - Connectionist model of sentence comprehension
    ... more
    Language and ability to communicate are thanks to their importance to humanity one of the most discussed phenomena in scientific field. The aim of detailed analysis of their nature lead to the creation of separate scientific disciplines focused on their specific attributes. Behavioural methods were the key explanations in certain particular signs although because of the difficulties of their design and often ethical restrictions they cannot be applicable in some cases of research. Computational models bring into this manner new direction of their investigation. In this work we aimed to create biologically plausible model of language comprehension based on a principle of self-organization and its comparison with model using supervised learning in sentence processing component. We used Self-Organizing Map for the representation of the meaning of the sentences. As a sentence processing component was chosen Merge Self-Organizing Map. On the other hand the reference model used Simple Recurrent Network. These components were trained to map the meaning of the sentences to the their text representation even after processed part of these sentences. We tested these models to predict the meaning of the sentences after the particular words were presented to them. We also scrutinized how well they can reconstruct the specific elements representing meaning of particular sentence. The results suggest, the model based on self-organization principle can sufficiently predict the meaning of the sentence after presenting each word in comparison with the reference model. In the prediction task the self-organization based model performed even better then the other model. Thus, we came to the conclusion that biologically plausible model based on self-organization principle can compete with classical models using supervised learning and therefore it can be appropriate alternative in simulation of cognitive processes related to the tasks of sentence comprehension.

  • Mgr. Marek Osrman - Constrained by convention: Emotional effect of microtonality on professional musicians and non-musicians
    ... more
    The goal of this thesis is to identify and explore the differences in perception of non-standard (microtonal) chords by musicians and non-musicians on GEMS (Geneva Emotional Music Scale). On the theoretical side, the work contains overview of acoustics, psychoacoustics and tuning systems that are beyond the standard equal temperament. Experimental part represents the results of the study with 35 audio stimuli (chords based mainly on Just Intonation tuning) which were evaluated by participants on 9 emotional scales (Joyful Activation, Sentiment, Peacefulness, Power, Sadness, Tenderness, Tension, Transcendence and Wonder). The experiment conducted on the sample set of 13 participants (6 musicians and 7 non-musicians) indicated statistically significant differences in perceiving certain emotions. Findings from Mann-Whitney U Test showed differences in emotions Joyful Activation (p = 0.18) and Wonder (p = 0.21). The emotion correlation matrices for both groups yielded major differences (Δp > 0.5) in pairs of Energy x Sadness and Wonder x Sadness, with both cases leading to negative correlation in musicians. Possible explanations of these findings are elaborated in the discussion.

    Keywords: psychoacoustics, emotions, microtonality, xenharmony
  • Mgr. Michal Páleš - Sentiment Analysis with Named Entity Recognition from Internet Articles
    ... more
    With the dawn of the Internet, came the time, where almost everybody has access to almost non depletable amount of information. But with this, there are people, or groups of people that create the content of the web itself. Such information can, but not necessarily have to, be true and so I decided to research this problem through several existing approaches. My goal here is, to create an application, that is able to download data from online sources, like big newspaper sites, quantify the sentiment towards specific named entities and compare the results between the different online sources. For this purpose, we researched previous works and approaches to “Sentiment Analysis” and “Named Entity Recognition”. Thereafter, we selected, from the previous works such approaches, that we deemed important and interesting to work with. Afterwards we implemented and tested the aforementioned network architectures. After implementation, we experimented with the resulting scripts, by changing the network hyper-parameters, word level embeddings as well as the layers of the networks. Then, we proceeded to compare the resulting success rates. Additionally, we developed an application capable of downloading “live” data from news agency. Finally, we discussed several other possible improvements, which could potentially improve or ease the work on similar challenge.

  • Mgr. Lulu Považan - Decision Making Performance, Satisfaction and Stability in the Context of Adult ADHD
    ... more
    Research shows that children with ADHD have various cognitive difficulties through their life, from the major symptoms of attention deficits, impulsivity and often hyperactivity, to comorbidity with learning disorders and mental disorders like anxiety or depression. These children once become adults and the struggles continue. This master thesis addresses decision-making of adults with ADHD, specifically their intuitive versus analytical solutions to decision-making problems, satisfaction with choice and choice stability. We revise the theoretical and diagnostic background of ADHD, neurological concepts of ADHD, and current research on decision making in ADHD adults. In our research, the participants were asked to make three pairs of complex choices and to solve three cognitive biases-related decision-making problems (confirmation bias, attribution error and conjunction fallacy). Afterwards, they rated their satisfaction, willingness to change their choice, and subjective decision-making style. We compared two samples of participants (N = 76) - general population and ADHD adults - balanced in gender, age, education level and study discipline. We hypothesized that there would be differences on the side of ADHD adults, in means of lower choice satisfaction and stability, and more prevalent intuitive decision-making, therefore greater susceptibility to cognitive biases. The two samples did not differ in decision-making performance - thus, ADHD adults were not more prone to the three cognitive biases. Yet, they were constantly less satisfied with their choices and more willing to change them. However, these results were significant only in two out of six cases. This indicates that the differences between general population and people with ADHD in choice satisfaction and stability may depend on the context - for instance, what are people deciding about and how many attributes do they evaluate. Our findings shed light on ADHD adults' decision-making, while empirical evidence on this topic has been missing so far. Besides a base for further research, it might be beneficial for broadening the ADHD neurocognitive profiles.

  • Mgr. Martina Simanová - Rola lingvistických markerov, expertízy a doménovej špecifickosti v komunikácii neistoty
    ... more
    Uncertainty communication and judgment and decision-making form everyday part of our lives. Empirical findings show the influence of linguistic marker of uncertainty source and speaker´s level of expertise on hearer´s judgments towards speaker and aspects of the message as well as on decisions about hearer´s future actions. Based on the effort to maximize effectiveness of uncertainty communication and improvement of judgement and decision-making, we realised research aimed to experimentally prove, if within two different domains (health, finances) and two different levels of uncertainty (higher, lower) linguistic marker (internal, external) and speaker´s expertise (expert, novice) influence the judgment and decision-making of the recipient. Our work was an extended replica of M. Juanchich et al. (2017) experiment 4 with 2x2x2x2 mixed factor design. Independent variables linguistic marker and level of expertise were manipulated between-subject and variables domain and level of uncertainty within-subject. We investigated their influence on judgment of knowledge, wrongness, blame, support and encouragement of speaker, reliability and intuitiveness of their statements and tendency to act of hearers. Based on original study, the expert will be in dependent variables rated higher using internal linguistic marker and novice vice versa. The findings of our work confirmed the impact of the linguistic marker, the level of expertise and even the level of uncertainty on the judgment and decision-making of the hearer, but they did not replicate the findings of the original study. Moreover, differences in effects of independent variables by domain were observed. As our research was the pilot one, based on its expanded design and focus and the results are not consistent, we are not allowing us to generalized recommendations for practice so far. But we recommend continuing in this research by its replications for the subsequent possible use of findings in setting up the communication standards for experts in the domains working with risk.

  • Mgr. Petra Šlahorová - Behavioural and electrophysiological characteristics of cognitive control
    ... more
    This thesis examined behavioural and electrophysiological characteristics of cognitive control in relation to advancing age and clinical condition. The total number of 90 subjects consisted of 2 healthy (young and old) and three clinical groups (patients in early stage of Parkinson‘s disease; in later stage with mild cognitive impairment; and patients with amnestic mild cognitive impairment with high risk of progression to Alzheimer‘s disease). All subjects performed AX-Continuous Performance Task (AX-CPT) which is usually applied to reveal deficits in context processing/maintenance and response execution. The results indicated preserved accuracy but slower performance in the task in all older groups. Even though we identified a number of cue-, delay-, and target-related effects across conditions we lacked to capture significant differences among most of the groups. Therefore, to provide more detailed explanation of underlying mechanisms, further research on larger samples is needed, with focus also on ERP latencies.

    Keywords: AX-CPT, cognitive control, cognitive impairment

2017

  • Mgr. Jakub Benko - Influence of interoception on subjective time perception during isometric contraction
    ... more
    The thesis presents results of research on interoceptive sensitivity with focus on musculoskeletal afferentation. We devised an experiment, in which the upper limb vessels of participant are occluded during time perception task. We found that occlusion had significant effect on time perception during this task. Besides main goal, we found out that conscious interoceptive sensitivity is dependent on systolic blood pressure

  • Mgr. Peter Buzáš - Influence of Response Type on P300 Elicited in Oddball Task
    ... more
    The aim of this study was to find out how does the response type influence the P300 evoked potential in the oddball task. For this purpose we compared three ways of responding. Two traditionally used – pressing a button to target stimuli and mental counting of stimuli and one new way – pressing a button to every stimulus, one to target stimuli and one to standard and distractor stimuli. We compared these three methods of responding with auditory and visual stimuli. We found statistically highly significant effect of response type on P300 evoked potentials. Based on our results we conclude that this new type of responding may be used in the future if enough research attention is given to it.

  • Mgr. Nikola Ondríková - Autism Subtypes in Context of Current Theories
    ... more
    Autism is a neurodevelopmental disorder usually defined with a triad of manifestations – 1) impaired social interaction and 2) communication, and 3) repetitive behavior. The question what causes the symptomatology remains unanswered. The thesis aims to describe autism in a context of traditional theories of autism. Bayesian Brain Hypothesis is currently predominant theory of autism in cognitive science. The theory offers a roof for most of the other theories. Bayesian Brain Hypothesis understands the brain as a probabilistic machine which uses its general model to generate predictions. Current sensory input is tested against predictions or priors, and general internal model updates its beliefs about causes of the input. Predictions of the general model are inaccurate in autism. Bayesian perspective is promising account for autism subtyping due to possibility of modelling subtypes using various parameters. Future computational models could help to better understand the nature of the differences within autistic group and explain inconsistencies in biological research. From previous findings, it seems like there are two main subtypes - "less severe" and "more severe". In this study, we derive clusters from behavioral data employing cluster validation and hierarchical cluster analysis with model-based clustering. Our dataset consists of children's behavior description from their parent's standardized Interview (ADI-R) and clinician's observation (ADOS). Sample involves 217 autistic children, 13,36% girls 7,2 years old in average (SD =4,44) and 86.64% boys 7,5 (SD = 4,64) years old in average. Our results are in line with earlier findings and suggest the existence of two main clusters based on onset of developmental anomalies and gesture communication. We present a possibility of association between severity of autism, ability to form representations and motivation to social stimuli.

  • Mgr. Karina Rerichová - The Role of Steroid Hormones in Human Social Behaviors
    ... more
    The self-other distinction is an ability to distinguish self from other related mental representation, to take another's perspective and compute their likely emotional state which is an important socio-cognitive ability necessary for successful human interactions. These social interactions can improve our reputation and social status, which leads to better access to limited resources. Internal physiological states of the human body, including the concentration of hormones, can influence status-seeking behavior and empathy. For example, according to the dual-hormone hypothesis, testosterone increases behaviors related to status-seeking, such as overbidding in auctions or decrease of self-reported empathy mainly among individuals with the low levels of cortisol. However, testosterone effects on empathy and its interactions with cortisol levels have not been systematically examined. We addressed these open questions in a double-blind, randomized, between-subject design study, in which 40 healthy male participants received a single dose of 150 mg topical testosterone or placebo. Thereafter, they were subjected to a nonstressful condition (Warm-water test) to control for low cortisol baseline levels. Subsequently, they completed the self-other distinction tasks (Reading the Mind in the Eye Test, Emotion Recognition Task, Perspective-taking Task and Imitation-inhibition Task) that assess the self-other distinction ability. Saliva samples for testosterone, cortisol, as well as continuous blood pressure measures, were taken repeatedly throughout the procedure of the experiment. The results showed participants with the high baseline level of cortisol had higher levels of baseline testosterone. Moreover, participants who received testosterone, were more accurate in recognition of subtle emotional expressions.

    Keywords: testosterone, cortisol, dual-hormone hypothesis, self-other distinction
  • Mgr. Lukáš Rückschloss - Computational model of memory consolidation
    ... more
    The goal of this thesis is to design and implement a computational model of memory consolidation. The consolidation is a process of stabilization of short-term memories stored in hippocampus into long-term memories stored in neocortex. To do this we need to look at the function of the short-term hippocampal memory and long-term neocortical memory and the way they cooperate in storing the memories. With this in mind, we aimed for a designed and implemented model capable of consolidating memories. To do this, we used two self-organizing maps, one being the hippocampus, the other neocortex, and wast knowledge gained from research done in neurobiology to create model capable of transforming episodic data into short-term memories stored in hippocampus that are later transformed into long-term memories stored in neocortex, taking the emotionality and recency of the episodic memories into account when consolidating them.

  • Mgr. Matej Sedláček - Relationship between Musical Abilities and Science, Technology, Engineering and Mathematics in Elementary and Middle School Students
    ... more
    This thesis explores the relationship between STEM (Science Technology, Engineering and Mathematics) and music from different aspects, including education, psychology, cognition and neuroscience. This relationship is based on transfer of learning, which represents a way of learning, when knowledge is obtained in one field and is available to utilize in some other field. This phenomenon may be approached on neural and behavioural levels and is divided into different categories, based on its context. The aim of theoretical part of this thesis is finding a common intersection of different scientific fields to reinforce interdisciplinarity. This study is carried out in collaboration with Slovak Academy of Sciences and OECD – CERI department, which is currently in the state of conductuding international testing of knowledge and creativity in order to explore the international differences between elementary school students. The experiment was in form of pre-/post-tests with intervention. Both series of tests consisted of 3 different types of tasks, derived from OECD PISA and TIMMS testing, exploring knowledge in STEM, VAM (Visual Arts and Music) from witch music knowledge was determined, and mathematical creativity. The data obtained from this testing on Slovak students narrowed the selection of subjects to two groups – eight grade students (n = 81) and third grade students (n = 162). The outcome serves to emphasize the possibility of achieving better motivation and understanding of natural sciences integrating music into educational process and bringing this opportunity to public educational system, which currently does not utilize the potential of this phenomenon. The results have shown only weak significant correlation and regression analysis shown a significant trend between music and STEM scores (r = 0.227; R2 = 0.051), music scores and divergent mathematical creativity (r = 0.184; R2 = 0.034) on third-graders data and even negative non-linear significant correlations and trends in regression between STEM scores and the amount of two types of music lesson activity - active music engagement (τ = -0.187; rs = -0.251; R2 = 0.073) and standard activity (τ = -0.265; rs = -0.355; R2 = 0.136). This finding demonstrates relationship between the two fields at least to some extent, yet the nature of it is quite disputable.

top

2016

  • Mgr. Eugénia Belková - Kritické myslenie ako prepojenie vyučovania matematiky a percepcie okolitého sveta
    ... more
    Critical thinking could be considered an investigating tool for the information age we live in. As educational system should prepare students for their future, critical thinking could be one of the main educational goals. One of the ways of teaching students to think critically is teaching mathematics towards the development of divergent thinking and including real-life problems into education. The goal of our research is to examine students’ relationship to mathematics. We examined if this relationship could be changed to a more positive one even after a short experience of students with alternative lessons designed. Data was collected from high school students (2nd and 3rd year) in a form of interviews and questionnaires that were handed to students before and after our lessons.

  • Mgr. Adam Bilisics - Intelligent semantic matcher for Obziva.sk
    ... more
    The problem of intelligent data search on web portals poses a challenge for cognitive science inspired machine learning methods based on semantic relationships between words. Obziva.sk is a service offer-demand portal, which contains an interestingly large set of live, real-world data as well as maintains a busy user traffic allowing successful experimentation with its users. In this paper we present our own solution to the search problem by creating intelligent semantic matcher that provides matching offers for the users’ demands, using the most recent methodologies such as machine learning and semantics. After creating the network of words’ relationships by the supervised learning function, the results of testing showed that our matcher is capable of providing users the correct search results with 47% accuracy.

  • Mgr. Martina Chraščová - Androgénová aktivita a kognitívne aspekty problémového správania u detí s poruchami autistického spektra
    ... more
    Autism spectrum disorders (ASD) are a set of heterogeneous neurodevelopmental conditions characterized by early-onset difficulties in social communication, cognition and unusually restricted, repetitive behavior and interests. Children with ASD often present with co-occurring problem behavior (e.g. aggressive behavior and self-injury). ASD and problem behavior such as aggression exhibit an incidence that is male biased. As such, androgen activity can be considered a plausible biological risk factor for these disorders. Extreme male brain theory suggest that androgens, especially testosterone, might play a role in the complex etiology of ASD. However, there is insufficient information about the association between increased androgen activity and problem behavior in children with ASD. Purpose of our study was to investigate a relationship between actual plasmatic testosterone levels, sensitivity of androgen receptor (AR) and cognitive aspects of problem behavior in boys with ASD. The study sample consisted of 51 boys (ages 3-15) with ASD. In all children, parents completed Behaviour Problems Inventory (BPI-01) consisted from specific sub-scales: self-injury, stereotypic and aggressive behavior. Sensitivity of AR was determined by the number of CAG repeats in the first exon of the gene encoding the AR. Total plasmatic testosterone levels and a number of CAG repeats were determined in the venous blood samples. It was found no correlation between plasmatic testosterone levels, number of CAG repeats and any of the sub-scales of BPI-01. Further, it was found that the effect of testosterone on behavior is not determined by lower, medium and higher number of CAG repeats. Although in our study, there was found no significant association between biological parameters and problem behavior in children with ASD, our results reflect current find-ings in neurotypical population. In the light of cognitive science and interdisciplinary ap-proach we point to potential limitations of current research. Testosterone might not be the only hormone that plays a role in etiopathogenesis of problem behavior in ASD and we suggest to investigate mutual interaction between testosterone and other hormones/neurotransmitters (cortisol, serotonín) in future research. Activity of enzymes in-volved in testosterone metabolism (5 alpha reductase, aromatase) also influences androgen effect. Moreover the presence of blood brain barrier influences transport of testosterone into cerebrospinal fluid and levels of testosterone in periphery might not reflect the testos-terone effects in brain. Moreover, some cognitive theories suggest that a specific impulse, associated with a negative or fear memory in our brain can contribute to the etiology of problem behavior as well. Further investigation of complex androgen activity and the cog-nitive factors related to problem behavior might bring more clarification to the role of in-creased androgen activity in behavioral problems in ASD.

  • Mgr. Tomáš Lenč - Neural entrainment to musical rhythms in dyslexia
    ... more
    Across languages, dyslectic individuals have problems with the accurate neural representation of phonological aspects of speech. Recently proposed “temporal sampling framework” (TSF) suggests that this phonological deficit may arise from the reduced sensitivity to speech prosody and syllable structure. State of the art neural models of speech perception suggest that auditory system “samples” continuous speech by entraining (phase-locking) endogenous oscillatory activity to its spectrotemporal modulation patterns at different timescales. TSF proposes that dyslectics may exhibit atypical entrainment at slow frequencies (< 10 Hz), corresponding to syllable and prosodic structure of speech. This would explain the observed difficulties in perceiving speech rhythm and its acoustic correlates. Given the fact that speech and music both exhibit rhythmic metrical structure and neural oscillations seem to play an important role in processing of this structure, TSF predicts deficits also in tasks involving musical rhythm. Dyslectics indeed show difficulties in behavioral tasks requiring musical rhythm perception, sensorimotor synchronization, and also impaired neural entrainment to simple rhythmic stimuli. However, no study has investigated neural correlates of complex rhythm perception, which gives rise to representation of a metrical structure. This thesis attempts to build upon the suggestion of TSF that dyslectics are “in tune but out of time”. An overview of recent literature is provided with the focus on the role of oscillatory entrainment in processing of speech and music, and in particular rhythmic aspects of both domains. Empirical evidence supporting the hypothesis that dyslectics have difficulties with processing rhythmic aspects of speech and music is reviewed. Results of studies suggesting that these difficulties stem from a deficit in neural entrainment are presented. A small-sample experiment was carried out to examine neural entrainment to musical meter in adult dyslectics and control participants. While EEG was recorded, subjects listened to either metrically simple of complex rhythm. It was hypothesized that the overall magnitude of neural entrainment at meter related frequencies will be lower in dyslectics. This effect should be particularly manifested in the complex rhythm, as it places higher demands on the brain networks that are supposed to process temporal structure in sound. However, no supportive evidence was found for either hypothesis. The results are discussed with respect to recent neuroscientific theories of dyslexia and musical rhythm processing.

  • Mgr. Jakub Lipták - Meranie empatie vo vzťahu k postojom voči utečencom
    ... more
    The refugee crisis that hit Europe is a significant social phenomenon. This work is an attempt to react on the discussion elicited by this crisis. The negative attitudes of the majority of Slovak population towards the refugees were an appropriate object on closer research from the viewpoint of the connection of these attitudes to the empathy of people, since the empathy is perceived as a key factor of the approach towards the other people; also in the public discussion, the empathy was often being connected with the attitudes towards the refugees. These facts were a good enough reason for conducting a research with the research question, whether there is statistically significant correlation between the empathy and the attitudes towards the refugees. The work offers The Empathy Quotient Questionnaire from Simon Baron-Cohen localised for Slovak language by backward-translating the English original which is one the work’s contributions. The next one is a newly created questionnaire of the attitudes towards the refugees. After the theoretical specification of the problem, the results of the research are presented in the research part of the work. 163 respondents living in Slovakia took part of the research. The results give three following conclusions. First, there was not found any statistically significant correlation between the empathy and the attitudes towards the refugees – neither in case of men, nor in the case of women. Second, the higher empathy of women as of men was confirmed and this difference was statistically significant. Third, the attitudes towards the refugees were not found to be significantly different by sex of the participants. However, the structure of these attitudes was different among the sexes. This work is a contribution to the public discussion on the problem of refugees and offers the options for the next research.

  • Mgr. Matúš Marton - Analýza mikrostavov EEG počas úlohy na udržanie zrakovej informácie v krátkodobej pamäti: porovnanie existujúcich nástrojov
    ... more
    Elektroencephalography (EEG) excels between other neuroimaging methods by its high temporal resolution. Important information about ongoing neuronal activity can be provided by analysis of dynamic changes in topographies of EEG signal, by means of so called microstates analysis, also designated as „atoms of thoughts“, which represent short periods of stable activations of synchronized neuronal networks. Statistical analysis of temporal features (like onset, offset or duration) between groups and conditions then allows to precisely assess the timing of cognitive processes. In this thesis we introduce principles of EEG data segmenatation into microstates and compare existing tools (software) enabling such analysis. We applied both applications (separate software Cartool and program Ragu implemented in Matlab) for analysis of data gathered during the task for retention of visual information in the short-term memory with the goal to find the difference in microstates with regards to performance of subjects in this task. We have found that both methods are extremely sensitive for little changes of initial parameters involved in the analysis. Moreover, both methods are focusing on quantitative analysis of differencies between groups (or conditions) and statistical evaluation of qualitative changes (apperance of different microstates between groups) is not possible. Despite of these difficulties we have succeed, in certain level, to show the difference between the group with higher and lower performnce by the means of microstate analysis.

  • Mgr. Milan Mitka - Practical approaches to ERP analysis in the context of visual short-term memory
    ... more
    We provide an overview of practical approaches to the analysis of even-related potentials (ERP) and demonstrate the use of such methods in the context of visual short-term memory (VSTM). Inter-individual differences in humans performing short-term memory tasks can be attributed to different factors including different stages of the memory process, namely encoding, retention (or maintenance), resistance to interference, and recall. We analyse which part of these stages is the strongest predictor of performance measured by response accuracy using data from a short-term memory task, in which 28 subjects shortly observed a visual target stimulus which was then covered with a mask for three seconds and had to be retained in short-term memory. Then, they were presented with a probe stimulus and had to judge whether it was rotated clckwise or counter-clockwise with respect to the target, and indicate their decision by a key press. First, we introduce the theory in a methodical fashion, starting with a global overview of memory (history, models, experiments), later focusing on the specific type of memory that we set out to explore, followed by a general description of electroencephalography (EEG) and event-related potentials (ERPs), including recent findings and questions that are still unresolved, such as the origin of ERPs. At the end of chapter one, we cover time-domain analysis and statistical methods developed to make reliable inferences based on electrophysiological data, which contain many observations. Second, we present the methodology and results of our exploratory analysis. We begin with a brief description of data collection and pre-processing, explain the sensor-level analysis from start to finish (data importing, time-locked averaging, plotting, statistical design, multiple comparisons correction). The main finding is a consistent area of strong correlation (r > 0.5, p < 0.01) between the response accuracy and the amplitude of the respective ERPs in the time window between 376 and 552 ms after the onset of the target stimulus, i.e. during the late encoding stage and extending shortly into the period of maintenance, over the medial part of the frontal lobe. Finally, we attempt to perform a source-level analysis (volume conduction model, forward and inverse modelling) in order to obtain an approximation of the neural sources of signals collected outside of the head. The limitations of our procedure are discussed at length in the final chapter along with concluding statements.

  • Mgr. Juraj Štancel - Multiagentový model vplyvu priestorovej štruktúry populácie na emergenciu kooperácie
    ... more
    Cooperation is a fundamental factor in many systems, including human society. Game theory has become a powerful framework for investigation of this social behaviour. Commonly well known Prisoner’s dilemma is the game that attracted the most of attention. In this game, two players have to decide if they are going to cooperate or not. Players are rewarded with R = b 􀀀 c in mutual cooperation, where b represents benefits to the recipient and c represents cost that incurs to the cooperator. Otherwise, in mutual defection, both players get P = 0. Unilateral cooperation results in profit T = b for defector and S = 􀀀c for cooperator. The ranking of the four payoff values is T > R > P > S. However, it is often difficult to assess the proper ranking of the payoffs for different behavioural patterns in real life situations. This has led to a considerable gap between theory and experimental evidence, and to an increasing discomfort with the Prisoner’s Dilemma as the only model to discuss cooperation. In order to investigate the cooperative behaviour in every possible situation, we prepared our model which consists of a square lattice 10 x 10 in the form of a torus where every site is occupied by a single individual. Each agent engages in a pairwise interaction with one of the agents within Moore neighbourhood. They play one of the five social games namely the Prisoner’s Dilemma (T > R > P > S), the assurance game (R > T > P > S), the chicken game (T > R > S > P), the privileged game (R > T > S > P) and the altruist’s dilemma (T > P > R > S), based on Heckathorn’s definitions. Results suggest that initial density of cooperation in population has a big influence in Prisoner’s Dilemma. Similarly, beyond certain threshold of costs parasites persist in the population. The ability to form clusters enables cooperators to persist, that is same as in assurance game. The ability to give rise to compact clusters lacks in the chicken game, on the other hand there exist filament-like clusters and cooperative individuals. The altruist’s dilemma strongly favors defectors and, to the contrary, the privileged game favors cooperative behaviour.

    Keywords: multi-agent model, cooperation, game theory
  • Mgr. Ing. Matúš Tuna - Deep reinforcement learning for computer games
    ... more
    The topic of this thesis are the recent advancements in the fields of reinforcement learning and deep learning. We focus on the Deep Q-learning algorithm that was previously shown to be capable of learning effective control policies from a high-dimensional representation of the environment in the domain of simple computer games. The main contributions of this thesis are the review and the analysis of the current research on the topic of the neural network based reinforcement learning algorithms and their potential applications, as well as the replication of some results from the Deep Q-learning literature. Based on the review and the analysis of the relevant research we proposed several methods based on the current deep learning research that could be used to extend and to improve deep reinforcement learning algorithms like Deep Q-learning so that this algorithm could “emulate” some of the core cognitive competences of human and animal minds that are necessary for artificial agents allowing them to act as competent agents in the environment. We also managed to perform a limited number of experiments with the Deep Q-learning algorithm based on our own implementation. We were able to replicate some results from the key Deep Q-learning literature and also improve these results by utilizing a different learning algorithm and weight initialization techniques compared to the original literature.

    Keywords: reinforcement learning, deep learning, neural networks, games.
  • Mgr. Lenka Valuš - Testing the cultural cognition theory of risk perception
    ... more
    Decades of research sought to find out how people form positions on risks they encounter. According to the Cultural cognition theory, individuals deal with risk-related information in a manner that mirrors and reinforces their cultural worldview. In our between-subject study (N = 339) we examined cognitive processes through which cultural values might affect evaluation of threats and benefits. Experimental manipulation was based on providing opposing arguments, either anonymous or ascribed to specific authors. We identified a general predisposition to be or not to be afraid among our Slovak sample, and supported the assumption “people are afraid of thing they are unfamiliar with”. In addition, changes in risk attitudes have been subject to the proximity of cultural preferences of the participants and the authors. We discuss our findings within an interdisciplinary context, and specify corresponding implications – ranging from debate on human rationality, through “science of science communication”, to public policy interventions.

2015

  • Mgr. Vladimír Barus - Author's Expertise, Certainty in Expression, and Comprehensibility of Text as Factors in Forming of Epistemic Trust
    ... more
    Living in the Information Age that followed the Digital Revolution provides us with at least as many challenges as it does with benefits. Every day we are flooded with "junk" information, as the society produces more information than necessary for normal functioning, with most information of low quality (e-mail spam is a typical example, with conspiracy theories and pseudoscience following right after). This is a new type of pollution, called information smog (Bezroukov, 2015). When seeking well-being for self and others using knowledge granted by a system of freely-accessible information, it seems imperative to be able to properly navigate it in terms of credibility of its contents. Epistemic trust is a potentially quantifiable attitude towards knowledge, a tendency to attribute information with a certain degree of truth value on a spectrum from complete belief to complete disbelief. In this thesis, we set out to investigate several factors that may bear influence on the process of attitude formation in terms of epistemic trust towards a piece of information (a text on a specific topic in our case). These are: expertise of the author (text author descriptions), expressed author‘s certainty regarding presented information, and comprehensibility. Using an 2x2x2 experimental framework, each participant was exposed to two excerpts with varying levels of the focal variables (e.g. an easily comprehensible text of an expert that expresses absolute certainty in his statements, an easily comprehensible text of a layman that expresses his certainty in percentages, etc.). Participants were randomly assigned to the experimental conditions, and rated the excerpts based on their perceived credibility and trust. We evaluated the amount of influence of individual factors on the decisions of participants regarding trust towards provided texts. The only independent variable that had significant influence was expressed certainty.

    Keywords: Epistemic trust. Epistemic vigilance. Scepticism. Credulity. Certainty.
  • Mgr. Andrej Čičmanský - Kognitívne aspekty hry Starcraft 2
    ... more
    The aim of our thesis was to determine whether variables importance during development of expertise is stable across all levels of expertise in playing Starcraft 2: Heart of the Swarm. Another aim was to investigate development of expertise for each playable race within the selected game. Last aim was to provide an overview of relevant studies dealing with cognitive research on computer games while focusing on research methodology. We focused mainly on studies about benefits of playing computer games and studies of expertise development. Theme of our study is cognitive research on computer games. Selected computer game belongs to category of computer real-time strategy games. Playing this type of games requires simultaneous and continuous actions, coordination and information processing. Data for our research was obtained from records of games, which contain all game actions performed by players. By measuring the variables representing attention, cognitive-motor and perceptual processes of players we have shown that the importance of these variables is not stable across all levels of expertise development. We extended research of expertise development by analysis of the expertise development among the playable races. To determine the importance of the variables, we used the classification through random forest. We compered our results with the previous research in this area.

  • Mgr. Richard Dinga - Decoding the Content of Spatial Working Memory: an fMRI Study
    ... more
    The goal of the thesis was to combine machine learning techniques and fMRI to decode the content of the spatial working memory from the recorded brain activity. Secondary goals were to localize brain areas in which pattern of activity carry information about the stimulus and evaluate nature of that representation and compare informative patterns of activities over time. Participants performed simple working memory task of remembering the position of a stimulus during which their brain activity was recorded by fMRI. We trained a support vector machine classifier to classify brain activities in respect to the position of the showed stimulus. We were able to decode in which of the 4 quadrants stimulus appeared with statistically significantly higher than chance level accuracy for 3 out of 4 recorded time points between stimulus presentation and response. Accuracy rate was statistically significantly higher if the classifier was tested on the data at the same time point as it was trained on. This suggests confounding signal in the data. We localized informative pattern of brain activity in the occipital lobe. This is compatible with the sensorimotor-recruitment models of working memory, which hypothesize that the working memory information is stored in the same areas that process the sensory information. We did not find informative patterns of activities in other areas that are often connected to the storage of working memory content in the literature – namely prefrontal and parietal cortex. Experimental design did not allow us to distinguish between the signal that is related to working memory processes and the signal that is related to different cognitive processes such as perception or attention.

  • Mgr. Miroslav Heriban - Effects of transcranial magnetic stimulation protocols on the neuroplasticity of the primary motor cortex
    ... more
    Our study, as a part of an ongoing project concerned with transcranial magnetic stimulation (TMS) as an alternative treatment of depression, was focused on testing of the most perspective continuous theta burst stimulation (cTBS) protocols, which are a subset of repetitive transcranial magnetic stimulation (rTMS) protocols, in their efficiency of inducing neuromodulatory changes in the primary motor cortex, concretely inhibitory effects. This would indicate their potential efficiency in inhibiting the dorsolateral prefrontal cortex, which is the brain structure that is currently being stimulated with conventional rTMS protocols as an alternative treatment of depression, but with a lower efficiency than other alternative treatments such as electroconvulsive theraphy (ECT), which has notable disadvantages including cognitive impairment. After an extensive review of literature and studies conducted with cTBS, we chose the standard 50 Hz cTBS and a relatively novel, but promising 30 Hz cTBS as the most suitable candidates for inducing inhibition in the primary motor cortex. We conceptualized a within-subject experimental design to assess their effects on the primary motor cortex. Our results showed that both tested cTBS protocols were not statistically significantly different in their effects on the primary motor cortex, which were slightly facilitatory rather than inhibitory. Furthermore, we found that the effects were intrapersonally as well as interpersonally variable, unstable over time and generally inconsistent. The cumulative effects of cTBS were also incoherent. Our findings have important implications on the TMS-depression project as well as the broad scientific community in TMS research, since the cTBS is widely considered as inhibitory. The limitations of our study include the potentially problematic assumption of the generalization of the primary motor cortex response on other brain structures, our general inexperience with TMS, the inherent variability of TMS, the author's general inexperience and other practical issues related to TMS.

    Keywords: transcranial magnetic stimulation, TMS, rTMS, theta burst stimulation, cTBS, primary motor cortex, depression
  • Mgr. Dalibor Jurášek - Cognitive science as a basis of integrative science
    ... more
    Contemporary science is predominantly based on a paradigm which assumes that everything is essentially material and governed by deterministic or random interactions. This approach has achieved successes in many fields, but faces substantial problems in three areas: the understanding of conscious phenomena; integration of knowledge into a holistic view; and answering the "big questions". In this work, I propose a method of first-person research based on advancing one's level of consciousness by training certain mental faculties. These include heightening the perceptiveness to inner experience, and developing the ability to retain a lucid, critical, self-aware consciousness in various altered states of consciousness (e. g. lucid dreaming). This method could help solve the limitations of the current paradigm: it would enable us to systematically observe and experiment with certain conscious phenomena; it could provide us insights that would help us unify our partial theories; and possibly grant us a qualitatively different type of understanding by reaching higher states of consciousness, such as mystical experiences. The observations and findings based on this method could be verified intersubjectively by researchers who have undergone the necessary training, and many of the insights could be verified and further specified by third-person research. A unification of first- and third-person methods would form a basis of integrative science which would help us better understand ourselves, nature and the world.

    Keywords: integrative science, first-person research, development of consciousness, materialism, paradigmatic change
  • Mgr. Matej Klobušník - Vplyv haptického vedenia na učenie rytmických sekvencií
    ... more
    In this thesis we are investigating the effect of proprioceptive feedback on motor learning with haptic guidance. Subjects are physically guided through a haptic interface producing the ideal motion providing them with proprioceptive information about the required position of their joints. We have constructed an electronic haptic keyboard. Fingers of the user are firmly attached to the keys, which can be moved programmatically together with the fingers. We have conducted three experiments, where musically naive subjects learned rhythmic sequences with varying temporal, ordinal or both properties. In the training phase, three conditions were used to present the subjects with the target sequences, namely haptic, audio, and audio-haptic. In haptic condition subject fingers were physically moved through required sequence, during audio they listened to the recording of the sequence and in audio-haptic both stimuli were combined. During testing phase subjects were required to reproduce the target sequences. The first experiment was testing the effect of haptic guidance on the ordinal property and also used as a validation for our device. The second experiment was targeting temporal property of rhythmic sequences where timing performance on more difficult sequences was measured. The third experiment was focused on rhythmic sequences with varying temporal and ordinal property. Results of the first experiment have shown that the ordinal property was learned significantly better during the audio-haptic condition. This is consistent with previous studies and this validates our device. Results from the second experiment confirm that the haptic audio condition was better for learning the temporal aspects. In the third experiment, better performance has been reached in both sequence properties. Our results support that this form of haptic guidance is more effective for teaching musically naive subjects to perform new rhythmic sequences, when compared with audio-only learning.

  • Mgr. Zdenko Kohút - Categorization, comparison and conventionalization in metaphor comprehension
    ... more
    There are two strong competing views on the nature of metaphor. First, represented by the class-inclusion theory proposes that metaphors are processed as categorizations. The other view, represented by the career of metaphor theory, claims that novel metaphors are processed as comparisons and only conventional metaphors are processed as categorizations. Based mainly on the career of metaphor theory and the relevance theory, we were able to describe the difference between categorization and comparison in metaphor comprehension in simple terms of number of properties shared by the concepts. We proposed that as the difference between the number of properties encoded by the abstract metaphorical and the target concept increases, the more likely will a metaphor be considered a categorization. Reversely, as the number of properties encoded by the abstract metaphorical concept approaches the number of properties encoded by the target, the more likely will a metaphor be considered a comparison. Hypotheses implied by this proposal were tested experimentally. Our results are consistent with our predictions concerning novel metaphors. Unfortunately, the variance in the data did not allow us to draw any conclusions about conventional metaphors. Additionally, we were concerned with neural aspects of figurative language processing in typical and non-typical populations and suggested a simple way to describe figurative language processing in schizophrenia and autism spectrum disorders consistent with our proposal.

  • Mgr. Jana Kováčiková - Distribution of information under conditions of uncertainty
    ... more
    In this work we attempted to bring anthropology and its methods back to the spotlight of cognitive science. We begin with an overview of possible influences which cultural and environmental factors could have on cognitive processes; e.g., gender-sensitive language accelerates gender identification in children. We explain concepts of the cognitive and cultural niches, cultural models, and socially distributed cognition in more detail to illuminate shared cognition. We pay particular attention to folkbiology as the intuitive method for gaining information about nature and myths as being incorrect but effective transmitters of information. We continue by discussing the position of anthropology among the disciplines of cognitive science and introduce the method which could represent an important link for closer cooperation between these disciplines - Cultural consensus theory (CCT), which obtains cultural truths about given topics from a group of people. CCT is used for ethnographic studies and studies of opinions or knowledge, e.g., what do Americans think are the most frequent causes of death in the US. Our aim was also to encourage cooperation between anthropological and psychological research by practical application of two methods. We have interviewed 120 hunters, shepherds, and animal activists about their beliefs about wolves. Then, we applied CCT model on data to divide the participants into groups, according to their shared knowledge and beliefs. Subsequently, we applied Cognitive diagnosis model (CDM) which divided the respondents according to the true or false knowledge they shared. CDM is primarily used for diagnostic feedback on proficiency levels in learners. We have investigated that the groups identified in analysis overlapped, which means that one group consistently believes in truth, while the other group consistently believes in false information. Further, we speculated that shared cultural agreements derived from the CCT assessment will correlate with groups of hunters, shepherds, and animal activists, due to their different perspectives, however, this hypothesis has proved to be unfounded.

    Keywords: culture, cultural anthropology, cognitive science, cultural consensus theory, cognitive diagnostic assessment, shared knowledge, folk biology, myth
  • Mgr. Igor Lacík - A computational model of intrinsic and extrinsic motivation for decision making and action-selection
    ... more
    ABSTRACT Sometimes a child might cry and can only be calmed by a mothers hug. Other times it has to be fed. When hungry or in distress, it will cry again. These actions have been selected and learnt thanks to the child being driven to perform different actions in various situations. In recent years, intrinsic and extrinsic motivation is being explored in order to build autonomous systems and research decision making in organisms. In this thesis we describe the implementation of a proposed model of decision making and action-selection based on intrinsic and extrinsic motivation. We delineate the abstractions of used motivational variables and neural network architectures enabling computer scientists to replicate our work on different models and at the same time offer an application that serves for preparation of test scenarios in order to assist cognitive science researchers to study motivation.

    Keywords: intrinsic & extrinsic motivation, reinforcement learning, neural networks, machine learning, simulation
  • Mgr. Mária Marušáková - Pragmatics and empathy in second language aptitude
    ... more
    This thesis focuses on pragmatical language aptitude and its possible relation to empathy. We formed a hypothesis stating our belief that individuals with greater ability to relate to others would also achieve higher pragmatical competence. Pragmatics is not a well-researched aspect of foreign language acquisition and before conducting a successful experiment we needed to locate a suitable pragmatical test designed for non-native speakers of English. Therefore we firstly provide an overview of existing tests for pragmatical language ability. After we conducted a pilot experiment to validate these tests, we chose the most acceptable one to employ in the subsequent study and proceeded to gather data of 69 Slovak native speakers. The resulting analysis revealed that there is indeed a significant correlation between individual’s empathy and pragmatical ability, but the scoring methods must be further validated.

    Keywords: pragmatics, empathy, language aptitude
  • Mgr. Samuel Sitáš - The role of mirror neurons in cognition
    ... more
    Mirror neurons are a class of brain cells fortuitously discovered in the premotor cortex of the macaque monkey that become active during both execution and observation of the same action. This straightforward property is what enabled these neurons to conquer the fields of cognitive science and become one of the most acclaimed discoveries from the turn of the century. Whether it was a result of rigorous research or a consequence of undue speculations is debatable. With this thesis I strive to open the debate and point the reader towards a critical reconsideration of what we currently know and think about the mirror neurons. I commence my efforts by providing a thorough introduction to the neurobiological background of the primate action observation--execution network, into which the mirror neurons are embedded, and which facilitates their function. I continue with a review of the methods and results of human mirror neuron studies that laid down the foundations for further incorporation of mirror theories into their respective fields, only to then turn the tables and reveal their problems as the critics point out. In the concluding part of the thesis, I myself assume the sceptical perspective and voice my own concerns regarding the plausibility of the mirror neuron on a conceptual level. I further provide the results of an original EEG study, with the intention to illustrate the limited explanatory power of this method. Finally, I present my own vision of a resonance system that approaches the mirror neurons rather reasonably.

  • Mgr. Ján Tóth - Modelling Early Sensorimotor Development with Intrinsic Motivation
    ... more
    AI researchers have long had the ambition to build a model possessing human-like traits of development, displaying the same drives and motivations. It is reasonable to search for answers in the early period of development, where both the observed subjects and the models are still tabula rasa, and thus easier to reason about. Early human development is believed to be quite well described, for instance by the constructivists. There is a variety of approaches for computational modelling of sensorimotor development, ranging from traditional cognitivist to the data grounded emergent systems. We chose to explore a particular embodiment inspired model, one where curiosity that is so natural for humans is factored in and serves as a drive to actively explore the world. We implemented a learning system motivated by curiosity and tested it on a suite of virtual worlds. We also provide possible interpretations of the model's behavior in concrete situations from the perspectives of curiosity and sensorimotor development theories.

    Keywords: autonomous systems, intrinsic motivation, developmental robotics, curiosity

top

2014

  • Mgr. Ján Cimra - Situácia tesnej prehry v teórii Flow
    ... more
    Flow is the mental state of operation in which a person performing an activity is fully immersed in a feeling of energized focus, full involvement, and enjoyment in the process of the activity. In essence, flow is characterized by complete absorption in what one does. Proposed by Mihály Csíkszentmihályi, this positive psychology concept has been widely referenced across a variety of fields. Near misses are widely believed to encourage future play, even in games of chance where the probability of winning remains constant from trial to trial. Some commercial gambling systems, particularly instant lotteries and slot machines, are contrived to ensure a higher frequency of near misses than would be expected by chance alone. Aim of this thesis is to incorporate situations known as Near miss to Csíkszentmihályi’s Flow theory. Theoretical proposal of our solution is followed by presentation of our experiments. The results of these experiments indicates positive effect of Near miss situation to one’s Flow.

  • Mgr. Matej Fandl - Multiagentový model šírenia sa emócií v dave
    ... more
    In this work, I designed and implemented a computational model of crowd emotion propagation, which is supposed to be used for real-time simulation of crowd dynamics in computer games and live animation. This model is a result of interdisciplinary approach to studying crowd dynamics. By analyzing existing models and studying contemporary findings from psychology and social psychology I build a theoretical basis and was able to choose certain assumptions, which I used during the design process. Model is performing on local interactions between agents only and is built up on emotional contagion and an assumption, that emotion is a product of cognition and that emotion shapes organisms action tendencies. Implemented model is presented in a simple application, where users input can cause fight or flight behavior of a large group of agents. Evaluation results have shown, that the model can easily be set up to model crowds with various readiness to act by mere change of the parameter values.

    Keywords: multi-agent systems, affective computing, emotions, emotional contagion, computational modelling
  • Mgr. Roman Koči - Osobnostné črty asociované s rôznymi motiváciami hrať voľnočasovú mobilnú hru Quiz House
    ... more
    The goal of our thesis was to find how personality traits affect the motivation to play casual mobile game Quiz House. Five factor model (openness to experience, consciousness, extroversion, agreeableness, neuroticism) was used to identify the personality by ten item personality inventory. Yee’s motivational model (social, achievement, immersion) and adjusted Yee’s test were used to find out player’s motivations. Based on 180 participants we found out that players who are open to experience are motivated by immersion, avoiding bore down. The conscientious people are motivated in this game by achievement, the feeling of satisfaction. Although we expected that extroverted and agreeable individuals would be motivated by social benefits, it was not true in our study and we did not find any connection between extroversion, agreeableness and socialization. Our results can be suitable starting point for other studies of various genres of mobile games. Mobile game designers can take advantage of these information and improve the motivational and engaging aspect of games.

  • Mgr. Matúš Konečný - Evaluating and experiencing framed outcomes: Feedback-related negativity investigation of description and ambiguity effects
    ... more
    Outcomes framed in terms of gains are usually considered more favorable than equal ones framed as losses. Differences in effects of framing on evaluations of certain and risky monetary deals were investigated in the present research. Additionally, event related potential (ERP) technique was used to examine analogy in evaluative processing between framed and objective gains and losses. Compared to gains, losses and ambiguity elicit a negative deflection in an ERP waveform, feedback-related negativity (FRN). Our experiment (N=17) consisted of 2 separate tasks. Participants first evaluated favorability of certain and risky outcomes. Following evaluations, outcomes were collected into final winnings in an EEG recording session task. All the outcomes resulted in objectively positive or neutral reward. Some outcomes, however, remained ambiguous and did not include information about the magnitude of the reward. No framing effect on evaluations of known outcomes was present. Responses to ambiguous outcomes, however, indicated significant framing effect. ERPs revealed no FRN differences related to gain and loss descriptions. On the other hand, ambiguity resulted in FRN independent from framing condition. Absence of framing effects in both FRN and evaluations between framed outcomes with known magnitudes could be indicative of outcome processing unbiased by description. The observed FRN associated with reward magnitude related ambiguity suggests an unexpectedly profound effect on outcome processing. Framing effect consistent responses to ambiguous outcomes dissociated from FRN suggest that other cognitive processes might be involved in behavior typical for framing effect.

  • Mgr. Milan Lajtoš - Nástroj na interaktívne experimentovanie s umelými neurónovými sieťami
    ... more
    This thesis deals with biologically-inspired computational model -- artificial neural networks. We argue that deep neural networks are a prospective computational model, and we point out that its application is heavily limited due to tools, which provide this functionality. We claim there is a need for new tools which can help us create and understand systems based on the neural networks. We define three principles, which should be adhered to when conceiving such environment, and we present a prototype which implements these principles.

  • Mgr. Marek Meňhart - Comprehension of indirect replies
    ... more
    Three experiments were conducted examining comprehension of indirect replies with focus on inferential processes underlying their comprehension. Indirect replies examined in this research were non-idiomatic excuses yielding particularized conversational implicatures (Holtgraves, 1999; Grice, 1975). Experiments were designed using eyetracking (Experiment 1) and self-paced reading (Experiment 2, Experiment 3) paradigms. In Experiment 1, we demonstrated that comprehension of indirect replies is a time-consuming process requiring increased cognitive effort. We used Experiment 2 to determine when during the reading of an indirect reply is the inferential process initialized, and whether it takes place during the reading of a reply or in a post hoc manner. Our results suggest that inferential process is present during the reading of an indirect reply, with reading pace gradually decreasing towards the end of an indirect reply. Experiment 3 did provide evidence that there is inter-individual variability in indirect speech comprehension. The results are discussed with respect to different notions of inferential approaches to indirect language comprehension (Gordon & Lakoff, 1975; Holtgraves, 2002; Searle, 1975).

    Keywords: communication, indirect replies, implicature, pragmatics
  • Mgr. Eliška Pätoprstá - Change Detection in Peripheral Vision Using Art Paintings and Snapshots
    ... more
    Peripheral field of vision provides a better concept and understanding of the environment in which the person is located. It gives us the ability to grasp the whole scene at once without the necessity to look at everything separately. One of its purposes is to catch one's attention. We were interested in the question whether is the processing of visual perception inside the peripheral field of vision influenced by the top-down information processing of the brain. We approached this question by testing the detection of changes on art paintings and on the snapshots of real life. The hypothesis was that there would be a significant difference in detecting the changes between these two types of visual inputs. There are scientific proofs that art has certain impact on human psyche, but is this impact so strong that it could influence the visual processing of information in the brain? Are people better prepared to notice a change that will happen in the environment that is more natural to them, or are they going to be distracted by the art and will not be able to notice the change at all? There is of course also possibility that there will be no difference at all. We have decided to split our research into two parts. During the first part, we created our own version of the previous experiments on the size and shape of peripheral field of vision that is involved in cognitive processes. We have found it to have elliptic shape of certain size that corresponded with settings and equipment in our laboratory. The finding of the size and shape of this ellipse was important for the second part of our experiment, where we measured the detection of changes itself, on digitised paintings and on snapshots. We mapped the acquired ellipse to the image and determined whether the change occurred inside or outside of it. This information was compared with the results of successfully/unsuccessfully detected changes. As we were interested in top-down processing, we determined that successfully detected change was the one when the participant answered what exactly changed on image, not that just a change occurred. We managed to find out that there really occurs certain different approach to paintings and snapshots. During the detection of changes on snapshots, the participants were acting rationally and if they did not notice the change, they acknowledged it straight away. On the contrary, when viewing the paintings, they often, probably unconsciously, started to fabricate the different versions of changes that could have happened.

    Keywords: eye-tracking, peripheral vision, foveal vision, change blindness, top-down processing
  • Mgr. Ondrej Spevak - The Influence of Concurrent Cognitive Tasks on Balance, Step Initiation and Gait in Healthy Young Adults
    ... more
    The aim of this thesis was to process data and investigate whether cognitive-motor interference will be shown. We have used data of 23 participants acquired during standing, initiating a step and walking under single and dual task conditions. Cognitive tasks loading spatial and verbal working memory were performed concurrently with motor tasks. Results of RM-ANOVA performed on the data support the view of motor function dependent on cognition, or to be more specific working memory. In addition, we have calculated effect sizes of both concurrent tasks during each motor task. Results of this study will further represent data of control group in the research of cognitive-motor interference in patients with multiple sclerosis.

    Keywords: Locomotion; Gait; Cognition; Dual tasking, Motor-cognitive interference
  • Mgr. Martin Turček - Sociálne vplyvy v rozhodovaní o vakcinácii
    ... more
    Goal of the thesis is to deepen our understanding of vaccination related decision making. Core aspect of the research is observing the dependence of vaccination willingness on emotional strength of a text, read by participants. Previous findings report minimal to none effect of vaccination messages on vaccination willingness. Possible effect of too strong emotionality of the messages stressed as a potential cause of this inefficiency. This are the foundations of inspiration of our research of influence of emotional strength of vaccination messages in textual form on participant’s willingness to get vaccinated. Experiment was conducted online. During the experiment participants read two types of texts. One about hypothetical disease A, one about hypothetical disease B. After reading they answered questions regadring their willingness to be vaccinated themselves and willingness to get their child vaccinated with vaccine againts the disease A or disease B.506 people participated in the experiment. In the texts we manipulated level of emotionality by changing the number and strength of emotionally charged adjectives in the texts. Results suggest, that effect of emotionality to a great extent depends on the type of text, in which emotionality is manipulated. Complexity of this influence seems to be dramatically higher than expected.

  • Mgr. Peter Vlk - Medziagentová komunikácia ako prostriedok dosahovania cieľov
    ... more
    In this paper we are comparing success rate of two groups of cognitive agents. Agents in the first group are capable of communication, agents in the second group are not capable of communication. We are testing out hypothesis that the group of agents capable of communication will show better results. Agents are placed into virtual environment. Each agent has a sensor which is allowing him to observe its surroundings and set of actions that he can carry out. Agents’ goal is to maximize reward obtained from the environment. At the beginning of simulation agents have no knowledge about strategy that which they should follow. They must learn it. Agents’ learning mechanisms are implemented in form of neural actor-critic.

2013

  • Mgr. Filip Tóth - Evolvovanie riadenia pohybu mobilného robota v neznámom prostredí
    ... more
    The research into mobile robotics has been receiving more and more attention recently. To study various approaches to control, computational cognitive science as well as artificial intelligence use various methods. A lot of effort is put into the development of autonomous intelligent behavior for robots, especially for use in an unknown environment. The goal of this work is the implementation of a learning mechanism for a mobile robot based on a mix of a genetic algorithm and reinforcement learning, which can based on sensor data, enable the movement of a robot in an unknown environment. The advantage of this system is that it enables the autonomous definition of its behavior. A simulation environment was created for this system in which a virtual robot can move. This model is based on an actual existing prototype. The project uses machine nature-inspired algorithms that equip the agent with a very important ability - the ability to learn. This creates a significant need to motivate the mechanism, from which we require certain knowledge, correctly. In the theoretical part we provide a walkthrough through some related work as well as used methods. The designed learning mechanism was successfully implemented in a simulated model and in the end we successfully used it in a real world environment. The designed algorithm allows a real robot to learn to navigate in an unknown environment and to learn the principles of collision avoidance. Designed algorithm operates in real time and it exhibits significant learning properties. We conclude that the used algorithm is appropriate for this type of task.

top

2012

  • Mgr. Michal Badura - Vizuálna pozornosť a akčné 3D videohry
    ... more
    We are living in the society, where the computers go along with us everywhere. We are using them not only for the work, but also for the entertainment and relax. Our opportunities to play computer games are almost everywhere: at home, in the bus on our smartphones, or even on the lunch break at work. This could lead to a question, if the computer games could have some influence on our cognitive functions. As we know from Hebb postulate, neurons that fire together, wire together. If we are doing the same task for a long time, we will definitely improve in that task. So if we play some type of games for the long time, it must have some influence on our cognitive mechanism. This work is focused on action 3D video games called „FPS (First-Person Shooter)“ and their influence on visual attention of the player. These games are very dynamic and the player need to be quick enough, if he want to succeed. We wanted to know, if playing this type of games could somehow improve the player`s visual attention. Our hypothesis was, that playing this type of games decrease reaction time needed for accomplishing some tasks depending on visual attention. To verify this hypothesis we used some standard test focused on visual attention. Specifically „attentional blink“, Stroop effect, „dual task“, visual search - feature search, visual search - conjuctive search and test on pure reaction time after some visual stimulus.

    Keywords: visual attention, video games, FPS, attentional blink, Stroop effect, dual task, feature search, conjuctive search, reaction time.
  • Mgr. Maroš Gálik - Agent-based model of emotional trajectories in online communities
    ... more
    This work proposes an agent-based model of emotional trajectories in chat-like communities. We study the behavior of “collective emotions” in these communities based on data from real online interactions. Using the available data, we model and predict the emotional behavior of single users in chat-like online communities and, based on that, we also derive group behavior of a collective of users. The reference data consists of publicly available IRC-conversations collected over a time-frame of several years. The only channel that is available in order to infer internal emotional states of users is the automatically annotated content of messages authored by users in the context of the temporal sequence of texts. We propose an agent-based model in which individual emotions are reduced to two internal states: valence and arousal. In the history of an agent's communication, we observe the agent's behavior for different events – the reactions in terms of messages in different emotional states. Using these observations, we can define an agent “personality”, so that each agent is a different individual or a member of a particular group. Here, an agent personality is represented by a list of n rules, where the units of a single rule correspond to expected conditions (settings) for choosing eliciting an emotional reaction in a specific situation. Using available data in order to initialize a model, we predict and investigate the “emotional trajectories” of the modeled values. The steps taken in this thesis are part of research towards modeling group behavior based on models of individual emotional response.

    Keywords: agent-based simulation, cyberemotions, collective emotions, emotion modeling, emotion prediction, affective control architectures
  • Mgr. Milan Halabuk - Testing the hierarchical neural network DBN in invariant object recognition
    ... more
    Purpose of this work is to test ability of Deep belief network in object classification problems. This model is deep network with two phase training. The first one is unsupervised pre-training based on stack of Restricted boltzman machines. The second one is fine-tuning which uses back-propagation of error derivates. In this work are used two main datasets for experiments. The first dataset is composed of 11 leaves classes. This dataset includes rotational, color, size and noise variability. The second dataset is composed of normal and abnormal faces. The goal was to find out relationship among different network topologies, dataset parameters and final testing error. I found out that unsupervised pre-training, which is used for weight initialization helps to achieve better classification performance than random weight initialization. Significance of this help depends on type of dataset. In faces dataset this help is bigger (6.8%) than in leaves dataset (1.1%). I also confirm that a higher number of neurons and hidden layers increased classification performance.

  • Mgr. Róbert Kotrík - Searching for a strategy of Monopoly game using cognitive and artificial intelligence approach
    ... more
    The purpose of this work is to find and describe a strategy in Monopoly game. The work goes through several levels. The first one is my own observation of participants on the principles of behavioural experiments, the second one is drilling information from literature about strategies of players followed by a synthesis of these two through their analysis. As a result of analysis, several outcomes of strategy are offered. All is completed with a strategy model for a human player. This model is then applied to an artificial agent-player using a rule-based system. Two optimisations of the strategy were applied. The first one used an approach of genetic algorithms which through evolution of certain parameters improved the strategy of an agent-player. The second optimisation was made with an aim to specify a decision-making process of an agent from a global perspective by means of a reinforcement learning approach using a neural network as a learning system to distinguish whether a decision of the strategy is correct or not. I created a Monopoly game simulator which enabled agent-players with various strategies to play the game. At the end of this work all four strategies were compared before their next optimisation. Finally, I present the results of my experiment and discuss them. It turned out that an agent-player which uses both genetic algorithm and reinforcement learning optimisation of the strategy is the best one.

  • Mgr. Barbora Mináriková - Jazyk a percepcia. Vplyv zahltenia jazykovej modality na rozlišovanie relačných kategórií
    ... more
    In our diploma thesis we have focused on theoretical overview of the impact of language on specific cognitive tasks. We have described several research papers that deal with similar issues, but we also seek to answer our research question - whether and how congestion affects verbal modality word repetition in our reasoning. For this purpose, we prepared an experiment in the form of a web application and tested it by respondents. In conclusion, we are analyzing the measured results as well as opportunities for further research in this area.

  • Mgr. Peter Pružinský - Interaction of emotions processing and olfactory perception
    ... more
    The main dimension of a multidimensional odor experience is valence and, thus, it is likely to influence mood, so that pleasant odors tend to induce positive moods, whereas unpleasant odors tend to induce negative moods. Odors produce effects on cognition and behavior that are similar to those produced by emotional stimuli. Odors can provoke changes in physiological parameters, such as heart rate or skin conductance, which are directly involved in the emotional response. These effects are usually interpreted as an interdependence of olfaction and emotion on overlapping neural systems. The influence of implicit emotions processing on the odor intensity, the next dimension of an odor, was also confirmed with the results of the presented behavioral study. The results showed a significant evidence of emotions processing influence on subjectively perceived intensity of an odor, suggesting that activation of amygdala correlates with increasing perception of intensity of an odor. The neural representation of affective space into more primary intensity and higher-order valence components is compatible with the notion that the amygdala supports low-level (intensity) stimulus-driven processing. These findings dovetail nicely with appraisal theories of human emotions, which emphasize that affective responses are not a simple reflection of the intrinsic quality (positive or negative) of a stimulus, but rather result from interactions among the person, the situational context and the stimulus. An idealized computational component model of appraisal theory of emotions, which formalizes this theory, was used to show how an odor theoretically fits into this concept.

  • Mgr. Ján Šilar - EEG mu rhythm desynchronisation: An electrophysiological evidence for mirror neurons activity
    ... more
    Mirror neurons are multimodal association neurons with motor properties in premotor and posterior parietal cortex that increase their activity and fire not only during action execution, but also while observing or hearing another individual performing the same or a similar action. Mirror neurons play important role in action recognition, imitation, empathy, and also in the theory of mind. Electrophysiological index of mirror neuron system activity is an EEG oscillation called mu rhythm. We created an EEG experiment with focus on mu rhythm as the indicator of mirror neuron system activity. The core research question was to find the highest mu power in relax condition and similar mu suppressions in self motor movement, and the same observed movement condition. Within the second research question we expected to find contralateral correlations in mu rhythm strength according to index finger side in nonrest conditions and hemispheric differences in mu power also in nonrest conditions. The results showed statistically significant suppression from baseline in mu oscillations over both hemispheres during nonrest conditions, supporting our expectations of the first research question. Results for our second question were statistically insignificant, what contradicts our expectation. However, from this finding we can assume that the mirror neuron system activity is not hemispherically differentiated for distinguishing left and right side movement, but rather approximate to distinguish and understand any kind of movement.

  • Mgr. Milan Uherčík - Význam priestorovej konceptualizácie pre znižovanie kognitívnej záťaže v grafickom užívateľskom prostredí
    ... more
    This paper analyses main problems which occur during cognitive load caused by human-computer interaction (HCI) in graphical user interface (GUI) and further interprets them from neuropsychological point of view. Our line of reasoning is based on an enhanced version of Baddeley’s model of working memory and the two-streams hypothesis. We point out the importance of adjusting GUI to the motor skills used in the HCI, as described in the embodied cognition theory, and list principles of spatial conceptualization which reduce cognitive stress in HCI.

  • Mgr. Michal Vince - Communication platform for agents in heterogeneous environment
    ... more
    The aim of this thesis is to implement a communication platform for agents in heterogeneous environment. In this thesis we propose an implementation of a multi-agent middleware, that is implemented according to FIPA specifications. To ensure agents are able to run in heterogeneous environment, we implemented this middleware in three popular languages, C++, Java and Python. The architecture of the platform is inspired by the FIPA Abstract Architecture and so are other components of the platform. We are using FIPA HTTP protocol for sending messages. We extended this protocol for the purpose of message polling. By using this HTTP protocol, which is REST-ful, we ensure that agents from our platform are able to communicate with agents from other FIPA compliant platforms. Messages are using language that is inspired the Speech Act Theory, FIPA ACL language. The platform has low system requirements and is able to run on different mobile devices. Because of the low system requirements and the use of FIPA ACL language this middleware is a suitable tool for cognitive science surveys.

  • Mgr. Gregor Zaťko - Invariantné rozpoznávanie objektov pomocou perceptrónu s konvolúciou
    ... more
    In this work I concern with an issue of processing of image data by biologically inspired techniques, namely the convolutional neural networks, which are based on structure of visual cortex in brain. They achieve quite excellent results in tasks of image pattern recognition. The aim of the work is to test a network of this type on natural objects (leaves) dataset and measure errors of various network configurations. Another aim is the comparison of the network on this concrete dataset with other pattern recognition technique, the Deep Belief Network (DBN), and approximate evaluation of difficulty of the dataset compared to other one, in this case MNIST (handwritten digits). As the result, it is shown that convolutional network quite significantly exceeds DBN on this task by its classification accuracy.

    Keywords: pattern recognition, image data processing, convolutional neural networks

2011

  • Mgr. Martin Bies - Analysis of constitutional reasoning
    ... more
    Legal reasoning is a process of making decisions, based on available facts and rules from the legal system. Constitutional reasoning creates an area for discussion about compliance of law to the basic set of values and principles - constitution. Such reasoning is based on specific methods and principles. Some of them are known and already described, other are consciously or even unconsciously generally used. I am proposing an analysis of these methods observed in empirical material, text of juridical decisions. Main subject of this work can be divided into 2 parts. First one uses concept of meta-argumentation in the application of formal framework to legal reasoning, how inconsistencies in system can be handled and how the decision about case can be made in such process. Second part is about the principles and methods, which are used for attacks on arguments. Legal argumentation uses natural language, which very often leaves open space for various types of attacks on arguments. Goal is an application of formal framework to the extent, which the problem of constitutional reasoning can allow and to identify places, where such approach can be applied and can help us to achieve better understanding of the process, and on the other hand places, where this approach fails and more importantly why it fails.

  • Mgr. Ján Kolek - Hlasom ovládaný kurzor myši
    ... more
    This paper presents a less traditional usage of digital signal processing, the voice controlled mouse cursor. The cursor is controlled with continuous pronouncing of vowels with possibility of instant change of direction (vowel) without requirement of silence. This work contains important theoretical background to theme and scheme and implementation of solution, how to control the mouse cursor with vowels. We used two approaches to recognize vowels. Linear predictive coding with usage of Durbin method and Discrete Fourier transform. We used these two methods, to calculate a spectral envelope of digital signal. This spectral envelope provides characteristic lines, which allow us the recognition. We implemented both approaches successfully. Paper provides detailed characterization of algorithms in C language and analysis of reached results.

    Keywords: sampling frequency, digital signal, formant, frequency domain, spectral envelope
  • Mgr. Martin Šabík - Vplyv spánkového deficitu na fenomén falošnej pamäti
    ... more
    Information stored in memory is not always accurate reflection of reality. Construction nature of memory enables faster and more effective retrieval of memories, but also leads to predispositions to errors. Whereas most studies support presumption that the consequence of sleep is increased occurrence of false memory, there are studies which deny this hypothesis. The objective of the present diploma thesis was to study the influence of sleep deprivation on the phenomenon of false memory and to confirm stated hypothesis. For the aim of this study we have programmed two experiments. First experiment was variation of Deese, Roediger-McDermott paradigm, the second experiment consisted of participants watching a video document featuring a real life situation and 12 hours later the participants were asked to decide whether given statement is correct or not. In first and in second experiment, there were found statistically significant differences in retrieving true and wrong information between the group of participants that slept and the group that did not sleep. In first experiment statistically significant differences were identified with selected level of significance even in retrieving key - false information. In second experiment the influence of sleep on retrieving false information showed less distinctive level of significance.

top